Table 2.

Characteristics of included studies on general pregnancy risk assessment.

Study	Country	Sample Size	Population	Model	Knowledge Source for AI	Type of Risk	Prediction Value of AI/AUC	Prediction Value of LR
Gorthi et al. (2009) [30]	India	240	A prospectively collected sample of pregnant women was used to assess the practical model. There were 200 training cases and 40 test cases.	Knowledge-based system	Literature	Risk classification	Training 93.4 %, test 82.5%	NA
Umoh and Nyoho (2015) [31]	Nigeria	30	Pregnant women (aged 25–40) were selected to test the theoretical model.	Intelligent fuzzy framework	Literature	High-risk pregnancy	Not assessed	NA
Fernandes et al. (2017) [32]	Brazil	1380	Retrospective validation of the documentation of pregnant women from the High-Risk Prenatal sector at MEJC was used to test the theoretical model.	Knowledge-based system	Predefined risk factors	Risk reclassification	Not assessed	NA
Chaminda and Sharmilan (2016) [33]	Sri Lanka	117	Pregnant women of different ages and lifestyles were used. (Unclear if retrospective or prospective.) There were 93 training cases and 24 testing cases.	Hybrid system: neuronal network and naïve Bayes algorithm	Predefined risk factors	Pregnancy risk assessment	ANN 80%, naïve Bayes 70%, novel hybrid approach 86%	NA
Moreira et al. (2018) [34]	Brazil, Portugal, Saudi Arabia, India, Russia	100	Parturient women diagnosed with a hypertensive disorder during pregnancy were used. All prospectively collected cases were used to test the model.	Artificial neural networks (ANN)	Patient’s history	Hypertensive disorder during pregnancy	Hybrid algorithm 93%	NA